RE: [SI-LIST] : "skin effect/depth calculation results"

About this list Date view Thread view Subject view Author view

From: [email protected]
Date: Fri Sep 01 2000 - 15:50:04 PDT


This is great from a theoretical stand point. But in reality,
the trace impedance, Zo, is fixed at a certain range by either
the driver or receiver. You will need to make the widest
possible trace while keeping the same impedance. This brings
us back to the original discussion.

George Tang

-----Original Message-----
From: C Deibele [mailto:[email protected]]
Sent: Friday, September 01, 2000 5:35 AM
To: SI-LIST
Subject: [SI-LIST] : "skin effect/depth calculation results"

Hi Everyone,

Yesterday we discussed so many different parts of skin effect and copper

losses. After reading so many good points brought up by everyone,
namely the
effect of the perimeter of the trace, as well as the characteristic
impedance of
the trace, I had really wanted to study the effects of increasing the
trace
width and the copper losses associated with increasing the trace width.

As a background, remember that

I(z)=Const/Zo exp^{-i Beta z}

where Beta is the wavenumber...
Const is a constant
z is a position on the z axis
and Zo is the characteristic impedance.

so dI/I = -dZo/Zo (for a given point in z)

==>Increasing Zo decreases the current flowing on the conductor.
==>decreasing the current on the conductor decreases the copper losses
on the
conductor.
==>Increasing Zo means the trace width gets *smaller*

So, the question I ask is: Which effect wins?

There is no simple answer to this question because the characteristic
impedance
is not a simple linear function of geometry. Take for instance coax,
which is
one of the easiest geometries/topologies to study:
Zo=60 ln(b/a)

If we restrict our attention to geometries where the frequency is high
enough
such that the skin depth, delta, is much greater than a
i.e. delta >>a, then quite simply,

dZo/Zo=-da(r)/[a(r) ln (b/a(r))]

so, inserting this relationship into the equation of I

dI/I=-dZo/Zo=da(r)/[a(r) ln (b/a(r))]

the denominator clearly shows that there is not a nice linear
relationship. Now stripline is also nonlinear....So, this says that
there is probably NEVER a nice and easy rule to decide whether making a
trace wider, or thicker will reduce copper losses.

so, the change in the characteristic impedance is not something simple
to
describe -- namely there are regimes where a simple change in radius may
be
easily described by the inverse nature, other regimes where the
logarithmic
properties are dominated...and a regime where the mix is the important
property.

So, I attempted to study one simple property. I used Ensemble v. 7 and
simulated a simple stripline geometry. copper upper conductor, copper
lower
conductor, copper trace, and perfect vacuum in the middle (We were
restricting
our discussions to copper loss...*NOT* dielectric loss)

So, the results were fairly convincing, for a stackup as follows

copper ground
50 mil vacuum
copper trace
50 mil vacuum
copper ground

All the traces were simulated were 1800 mils long
The reflection data was nearly identical (on the order of 0.005%) ...so
I won't
post it....
trace at 140 mils http://www.geocities.com/simulations00/140.jpg
trace at 130 mils http://www.geocities.com/simulations00/130.jpg

and examine the following:

trace at 14 mils http://www.geocities.com/simulations00/14.jpg
trace at 13 mils http://www.geocities.com/simulations00/13.jpg

These two sets of results clearly show that there is a discrepancy based
on the distributed nature of the transmission line circuit.

So, the first set shows that a wider trace has less loss.
the second set shows that the skinnier trace has less loss.

I do not want to get into any argument about semantics about the English
of the problem. This is to say, skin effect, or skin depth.

Craig Deibele

**** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected]. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****

**** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected]. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue May 08 2001 - 14:29:25 PDT